In early 2009, Mintel, a market research firm, conducted a study of consumer opinion in order to determine whether people were choosing green products during a recession.  Mintel reported that whereas green shopping had been on the rise prior to the recession, with the number of people saying that they regularly buy green products tripling from 2007 to 2008, this growth flattened during the recession.  In early 2009, 36% of Americans said that they often or regularly buy green; the same percentage as in the previous year.  Marcia Mogelonsky, senior research analyst at Mintel, explained that people’s priorities had changed.  They were not willing to pay higher prices for green during hard times. 

The following analysis of online traffic data suggests that the flattening of growth in green consumption during the recession may be less prominant in several areas of online retail.   The analysis, which was conducted in early November, 2009, compares websites that promote green products to their counterparts in conventional markets in terms of the amount of growth in online traffic they have attracted during the recession.  It examines four types of online retail:  comparison shopping, department stores that offer a broad range of home and personal products, clothing, and beauty products. 

 How the Data Was Gathered

The online traffic data for this study was gathered from Alexa, a service that ranks online traffic to specific websites using two variables, reach and page views.  Reach reflects the percentage of total global internet users that visit a particular site, and page views reflect the percentage of global page views that are attributed to a particular site.  Whereas reach gives us a snapshot of how many distinct users visit a site, page views are an indicator of how many pages these people view, which correlates with how much time each user spends on the site. 

According to Amazon, which owns Alexa, “Alexa gets its traffic data, including reach, page views and rank information, from a global panel of web users. The panel is used as a statistical sample of Internet usage to extrapolate overall traffic patterns and web usage information. The panel consists of Alexa Toolbar users and other sources [of] web usage information.”

Alexa data has been criticized for inaccuracy in the past because of the difficulty of extrapolating traffic patterns from the comparatively small number of users who install the Alexa toolbar.  Critics have argued that Alexa data inflates traffic rank of sites that are heavily visited by web-savvy users who have installed the Alexa toolbar.  See the Alexa site for further information on how Alexa calculates online traffic and a disclaimer regarding the accuracy of these calculations.  

In spite of these criticisms, inaccuracy of Alexa data is mitigated in this study for the following reasons.  First, Alexa no longer relies on their toolbar users alone as the sample from which traffic data are extrapolated.  Second, there is no apparent reason that some of the companies compared in this study would be more heavily used by web savvy visitors than others.   Third, because Alexa data is particularly susceptible to inaccuracy and manipulation for websites that receive less traffic, the sample used in this study is confined to relatively large retail sites, specifically sites that are ranked globally by Alexa as belonging to the top 150,000.   Most of the sites are ranked much higher than that.   A list of the traffic rank of each site in the sample for this study is available on the final page of this report.  Fourth, the study seeks to identify broad trends across a number of sites rather than draw conclusions about a particular site.  Inaccuracies that might apply to individual sites are less likely to affect trends across this broad range. 

The sites compared in this study were not randomly selected.  Rather, the sample was obtained through the following process.  First, a Google search was done on key terms associated with each category.  For example, when looking for eco friendly beauty products, the key term “organic cosmetics” was entered for a search.  Only sites that sell predominantly eco friendly products qualified for inclusion in our green retailers group.  Many of these sites also offered fair trade items.  As noted above, only sites that ranked in Alexa’s top 150,000 worldwide were selected.  Of the sites that met these criteria, up to five that appeared first among the non-paid results on the first ten pages of Google searches were chosen for each category.  There are fewer high traffic sites that meet the criteria for our green retailers, and therefore, only seven green companies qualified for inclusion in our sample.  Twenty conventional retailers qualified for inclusion. 

Because the selection was not random, conclusions regarding the broader population of online retail cannot be drawn.  Moreover, the figures reported below are imprecise because they are based on graphic representations of reach and page views rather than the original figures used by Alexa to create the graphs.  These graphs are available for anyone to view on the Alexa site.  In spite of these shortcomings, the results are intriguing because they convey a broad trend suggesting that growth in online traffic to green sites may have outpaced growth in traffic to conventional retail sites during the recession.  

Data Summaries and Analysis

The following charts show growth in online traffic across two year-long periods for each retail site.  The first period runs from January 1, 2008 to January 1, 2009.  The second period runs from November 1, 2008 to November 1, 2009.  Growth from one date to the same date the following year allows us to control for differences caused by different seasonal marketing strategies.  Alexa reports 3-month traffic averages at each point in time, so, for example, a January 1 data point actually represents the average online traffic to a site over a 3-month period preceding January 1.  

Tables 1 and 2 below show the number and percentage of retailers in each group (green v. conventional) that experienced growth in online traffic during the two year-long time spans.  Table 1 shows that 57% of green retailers in our sample grew in reach (percentage of global internet users visiting a site) during the first period (1/08-1/09), whereas 29% of the conventional retailers grew in reach during this period.  However, 40% of conventional retailers enjoyed a growth in page views, whereas only 29% of green retailers experienced growth in page views during the period. 

Although growth in page views without corresponding growth in reach many appear to be counterintuitive, many factors can influence page views other than reach.  For example, a company may decide to shift advertising dollars from forms of advertising that attract many casual visitors, a high proportion of whom quickly leave the site, to more limited, but targeted forms of advertising that attract a smaller number of seriously interested customers who spend more time on the site.

 Table 2 shows that over the second period (11/08-11/09), the percentage of green retailers that experienced growth in traffic outpaced the percentage of conventional retailers experiencing growth with respect to both reach and page views.

The following tables compare how much online traffic growth green and conventional retailers each experienced on average as a group.  Table 3 shows how much average reach growth each group of retailers experienced over each of the two periods.  From 1/08 to 1/09 the percentage of global internet users visiting the green retail sites grew by an average of .001887 percent, whereas the percentage of global internet users visiting the conventional retail sites dropped by .018492 percent.  From 11/08 to 11/09, the percentage of global users visiting the green sites increased by .000150, whereas the percentage of global users visiting the conventional sites grew by .001583. 

 Table 4 shows how much the gain and loss reported in Table 3 represents relative to the starting point for the companies.  This is expressed as percentage increases.  For example, if a company’s traffic reach in January, 2008 was .1 and it grew to .2 by January, 2009, the percentage increase in reach over that period would be 100% because its reach for the period doubled.  Table 4 shows that the average percentage increase for green retailers was 44% for 1/08-1/09 and 22% for 11/08-11/09.  Conventional retailers experienced an average 9% decline in reach over the first period and an average 12% increase in reach over the second period. 

The average percentage increase in reach over the second period was greater for the green retailers even though their average reach growth was lower relative to the conventional retailers.  The reason is that the green retailers are generally smaller companies.  Their traffic rank at the beginning of each period is significantly smaller on average than that of the conventional retailers.  Therefore, the same amount of growth in reach for each group will represent a greater percentage increase for the green companies relative to the conventional companies. 

Although reporting percentage increase in traffic growth has the advantage of giving us a more complete picture of how much companies have grown, one problem that arises is that if a company’s traffic rank approaches zero at the beginning of one of the periods for which growth is being charted, very small traffic increases translate into very large percentage increases, which distorts the average percentage increase for the group.  For example, if company A begins with .0000001 percent of global internet users in 1/08 and grows to .0000003 percent in 1/09, this is a 300% increase in reach growth, which would significantly raise the average percentage increase for the group to which this company belongs, presenting a distorted view of the group’s performance as a whole.  Moreover, working from Alexa charts makes precise readings of these data points impossible, and because tiny increments in growth translate into very large percentage increases, including them would greatly increase the likelihood of reporting significant errors and misleading results. 

Therefore, when a company’s traffic approached zero at the beginning of one of the periods used to chart growth, it was reported as “too small to identify” and was not included in growth calculations for that period.  This occurred with one green company and one conventional company when calculating reach growth for the first period.  It occurred for two green companies and two conventional companies when calculating page view growth during the first period, and it occurred for two green companies when calculating page view growth during the second period.

Among the companies reviewed in our sample, one business in our clothing category, Lulus, stood out as a true outlier in terms of percentage increase in traffic growth during the second period.  The company’s percentage increase in reach growth over this period was 200%, far greater than any other company.  Because this significantly skewed the overall average performance of the conventional companies over this period, Table 5 has been included, which excludes Lulus from the calculations.  Table 5 shows that the percentage increase in reach growth for the conventional group drops from 12% to 2% when Lulus is excluded.

Table 6 presents the overall average reach growth and average percentage increase in reach growth across both periods.  To find average overall reach growth, the data reported in Table 3 for each of the two periods was added together and divided by two for each group.  The same was done using the Table 4 data to calculate an overall percentage increase in reach growth for each group.  The result shows that the percentage of global internet users visiting the green companies grew by an average of .001019, whereas the percentage of global internet users visiting the conventional companies decreased by an average of .0016909.  This represented an average 33% increase for the green retailers and an average 3% decline for the conventional retailers.  Table 7 shows that when Lulus is excluded from the calculations for reasons explained above, the remaining conventional retailers experienced an average overall 7% decline over both periods.

That conventional retailers experienced an overall decline in reach growth, yet showed a slight increase in overall percentage reach growth may seem counterintuitive, so an explanation is in order.  To simplify, imagine that on day 1, company A had a reach value of .10 and on day 360, it had a reach value of .08.  Now imagine that on day 1, company B had a reach value of .01 and on day 360 had a reach value of .02.  The average change in reach value for these two companies over the period is calculated by adding the decline experienced by company A (-.02) and the growth experienced by company B (.01), then dividing the difference by two (-.01 divided by 2), which gives us -.005.  However, when we calculate % reach change over this period, we are determining how much reach value the companies gained or lost relative where they started at the beginning of the period.  In the above example, company A experienced a 20% decline in its reach value.  Company B experienced 100% growth in its reach value over the period.  To calculate the average % reach growth for both companies during the period, we would subtract company A's 20% decline from company B's 100% increase, and divide by 2, which yields an average growth of 40% for these companies over the period.  Although they would have experienced an average decline in reach value, their average % growth over the period would have increased.  

 Whereas Tables 3 through 7 present summaries of change in reach, Tables 8 through 12 present a similar set of summaries for page views.  Table 8 shows that both the green and conventional company groups experienced average growth in page views over both periods.  The percentage of global page views attributed to the green retail group grew by an average of .000058 during the first period and .000006 in the second period.  The percentage of global page views attributed to the conventional companies grew by an average .000307 during the first period and .000448 during the second period.  Table 9 shows that this represented an average percentage increase of 23% for the green retailers and 2% for the conventional retailers during the first period, and an average percentage increase of 18% for the green retailers and 41% for the conventional retailers during the second period. 

Although it appears that the conventional retailers performed better than the green retailers with respect to page view growth during the second period, this is misleading because Lulus was an even more extreme outlier with respect to percentage increase in page views than was the case with reach.  The company enjoyed a 700% growth in their share of the percentage of global page views from 11/08 to 11/09.  Table 10 shows that when Lulus is excluded from the calculations, the average percentage increase in page views growth for conventional retailers during this period drops from 41% to 6%, one third that of the green retailers.  

 Table 11 shows the results when page view growth is averaged for both periods.    The global percentage of page views grew by an average .0000032 for green companies and .000378 for conventional companies.  This represented a 21% increase for the green companies and a 22% increase for the conventional companies.  Again, much of this growth for the conventional companies is attributable to the extraordinary performance of one company.  Table 12 illustrates that when Lulus is taken out of the equation, overall page view growth for the conventional companies drops to 4% for the conventional companies in comparison to 21% for the green companies.

As noted above, green retailers included in our sample were smaller on average than conventional companies.  One possible explanation for the greater percentage increase in traffic growth for the green retailers is that they grew more relative to their initial starting point simply because they had lower traffic ranks to start with.  In other words, perhaps it is simply easier for smaller companies to chart growth because they have more room to grow in comparison to companies that already have acquired a large share of internet traffic.  To evaluate this explanation for the difference between the green and conventional groups, we tested for correlation between the Alexa traffic rank of all the companies tracked and growth in traffic during the recession.  The outcome is illustrated in Table 13, which shows no significant correlation in our data between traffic rank and traffic growth over our two periods. 

Correlation values are expressed on a scale of -1 to 1.  A score of 1 indicates a perfect correlation between two arrays of numbers, meaning that as the value of the numbers in one array increase, the numbers in the other array also increase.  A score of -1 indicates a perfect negative correlation.  In this case, as the numbers in one array increase, the numbers in the second array decrease.  A score of 0 indicates absolutely no correlation between two sets of numbers.  Table 13 shows that when our array of overall reach growth was compared to the traffic rank of the companies, there was a minor positive correlation.  Alexa’s traffic rank numbers are inversely related to their reach and page view percentages, meaning that as reach and page views increase for a company and their traffic rank improves, their traffic rank number declines.  The company that has the most traffic will be given a rank of number 1.  Therefore, the minor positive correlation between reach growth and traffic rank among the companies in our sample means that companies with lower traffic rank (higher numbers) were slightly more likely to also have higher reach growth numbers.  However, the opposite was true for page views.  Companies with lower traffic rank were slightly less likely to have higher growth in page views.  When percentage increase in reach and page view growth was considered, correlations were negligible.  Therefore, differences in recession-period growth between the green and conventional retailers in our sample do not appear to be attributable to differences in the traffic rank of the companies in the sample.
   

 Conclusions

As noted at the outset of this report, the results are merely suggestive.  The use of a small sample and lack of random sampling prevents us from drawing conclusions about the broader populations of green and conventional retail, and the method of using Alexa graphs to gather the data yields imprecise, and possibly inaccurate results.  Nevertheless, the graphs did appear to track a broad upward trend in online traffic for the green retailers in our sample relative to the conventional retailers, which is particularly intriguing in light of the conclusions from the Mintel report that consumer preference for green products had flattened during the recession. 

It is possible that this flattening effect is more characteristic of traditional consumers than those who engage in ecommerce.  Certain groups of green consumers are more likely to do their shopping online.  For example, according to an article entitled Green Consumer Demographics (eMarketer, 6/23/08),  young consumers are more likely than older consumers to buy green and use the internet.  Thirty-two percent of teens engage in ecommerce and forty-two percent of “green teens” (defined as “teens especially concerned and committed to environmental issues”) engage in ecommerce.  Perhaps populations such as these are more resistant to shopping price at the expense of environmental concern during an economic downturn.

Appendix:  Raw Data